Comment author: llb

Oops, sorry, bad title.
Should be "Error not raised with max_int+1 integer literal", or something like that.

Comment author: @damiendoligez

Note: for technical reasons, this is unlikely to be fixed in the near future.

Comment author: @xavierleroy

Let me document the "technical reasons".

1- Because "f-1" should not be parsed as "f(-1)", "-1073741824" is, lexically, the unary minus operator applied to "1073741824". So, we need to be able to represent "1073741824" (the opposite of min_int) in lexer tokens.

2- Still assuming a 32-bit platform, we really want to be able to write
"x land 0x4000_0000" (or similar bit-masking operations) and not be forced to write "x land (-0x4000_0000)".

The problem reported could possibly be solved by keeping integer literals as strings during lexing and parsing, pushing their resolution into integers deeper in the compiler. But that's a lot of changes for a minor benefit. I'm suspending this PR until someone comes up with a better idea.